Rare Copy Number Variations in Congenital Heart Disease Patients Identify Unique Genes in Left-Right Patterning

Rare copy number variations in congenital heart disease patients identify unique genes in left-right patterning

Khalid A. Fakhroa,b, Murim Choia,b, Stephanie M. Warec, John W. Belmontd, Jeffrey A. Towbinc, Richard P. Liftona,b,1, Mustafa K. Khokhaa,e,1, and Martina Bruecknera,e,1

aDepartment of Genetics, bThe Howard Hughes Medical Institute, and eDepartment of Pediatrics, Yale University School of Medicine, New Haven, CT 06520; cDepartment of Pediatrics, Cincinnati Children’s Hospital Medical Center, University of Cincinnati College of Medicine, Cincinnati, OH 45229; and dDepartment of Molecular and Human Genetics, Baylor College of Medicine, Houston, TX 77030

Contributed by Richard P. Lifton, January 3, 2011 (sent for review September 21, 2010) Dominant human genetic diseases that impair reproductive fitness embryos; a network of genes involved in the formation and and have high locus heterogeneity constitute a problem for gene function of the ciliated LR organizer that is conserved across all discovery because the usual criterion of finding more mutations vertebrates has been described; however mutations in identified in specific genes than expected by chance may require extremely genes account for less than 10% of affected Htx subjects (9–18). large populations. Heterotaxy (Htx), a congenital heart disease re- A major limitation in identifying causative genes in Htx is the sulting from abnormalities in left-right (LR) body patterning, has paucity of families segregating highly penetrant alleles, and the features suggesting that many cases fall into this category. In this high locus heterogeneity, which has limited the ability to map setting, appropriate model systems may provide a means to sup- disease loci. Because of the marked impairment in reproductive fi port implication of speci c genes. By high-resolution genotyping of fitness, some fraction of Htx could be caused by very rare, highly 262 Htx subjects and 991 controls, we identify a twofold excess of penetrant, dominant mutations. subjects with rare genic copy number variations in Htx (14.5% vs. − Although such mutations have historically been difficult to 7.4%, P = 1.5 × 10 4). Although 7 of 45 Htx copy number variations identify, recent advances have improved the ability to detect

were large chromosomal abnormalities, 38 smaller copy number MEDICAL SCIENCES variations altered a total of 61 genes, 22 of which had Xenopus these. For example, the use of quantitative interrogation of dense orthologs. In situ hybridization identified 7 of these 22 genes with sets of SNPs has dramatically improved the ability to detect small fi expression in the ciliated LR organizer (gastrocoel roof plate), a copy number variants (CNVs) (19, 20). The signi cance of such fi marked enrichment compared with 40 of 845 previously studied rare mutations can be dif cult to establish in the setting of high genes (sevenfold enrichment, P < 10−6). Morpholino knockdown locus heterogeneity, as is the case for Htx, where discovering a in Xenopus of Htx candidates demonstrated that five (NEK2, second hit in the same gene in a small cohort is unlikely. Alter- ROCK2, TGFBR2, GALNT11, and NUP188) strongly disrupted both natively, their significance can potentially be assessed in high- morphological LR development and expression of pitx2, a molecular throughput model systems. We have used Xenopus tropicalis, marker of LR patterning. These effects were specific, because 0 of motivated by a conserved mechanism of LR development and 13 control genes from rare Htx or control copy number variations prior use of this animal model system for robust cardiac and gut- produced significant LR abnormalities (P = 0.001). These findings looping assays. Moreover, Xenopus cardiac morphology is more identify genes not previously implicated in LR patterning. similar to human than fish-human similarity (e.g., the presence of atrial septation), and its relatively compact diploid genome (1.5 cardiac development | Xenopus tropicalis | embryo Gb) retains substantial synteny to human, simplifying the identification of orthologous genes (21–23). Additionally, the ability to ongenital heart disease (CHD) is the most common major produce large numbers of embryos and the absence of recent Cbirth defect, affecting ∼1% of live births, yet the cause of genome duplications facilitates screening by morpholino (MO) these lesions remains elusive. Although there is extensive evi- knockdown technology (23, 24). dence from epidemiologic, twin, and animal model studies supporting strong genetic contributions to CHD, only a small fraction Results of disease risk has been explained at the molecular level (1). Rare CNVs Are Overrepresented in Htx. We genotyped samples from Heterotaxy (Htx) is a severe form of CHD (2), in which normal 262 subjects with Htx, defined as any arrangement of organs left-right (LR) asymmetry is not properly established, leading to across the LR axis differing from complete situs solitus or malformation of any organ that is asymmetric along the LR axis complete situs inversus (Fig.1 and Table S1). Because there is (Fig. 1). During embryonic development, cardiac precursor cells evidence that human mutations in known LR patterning genes form a symmetric heart tube that undergoes rightward looping to can cause isolated transposition of the great arteries, we included form the geometric framework for the normal four-chambered patients with isolated transposition of the great arteries in our heart. Defects in looping result in a spectrum of complex CHD cohort (17, 25). in ∼90% of Htx patients (3). CHD associated with Htx still has relatively poor survival, despite surgical management. Studies in model systems have established a remarkably well- Author contributions: R.P.L., M.K.K., and M.B. designed research; K.A.F., M.C., M.K.K., and conserved genetic program governing patterning of the verte- M.B. performed research; S.M.W., J.W.B., and J.A.T. contributed new reagents; K.A.F., M.C., brate LR axis (4–6) and cardiac development. LR asymmetry is R.P.L., M.K.K., and M.B. analyzed data; and K.A.F., R.P.L., M.K.K., and M.B. wrote the paper. established during gastrulation at the node (LR organizer) via The authors declare no conflict of interest. dynein-driven, directional beating of cilia. Cilia beating results in Freely available online through the PNAS open access option. fl fl fl leftward ow of extraembryonic uid (nodal ow), which induces 1To whom correspondence may be addressed. E-mail: [email protected], nodal signaling and Pitx2 expression in the left lateral-plate [email protected], or [email protected]. mesoderm (7, 8). Abnormal LR development leads to a striking This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. and specific molecular and anatomic phenotype in all vertebrate 1073/pnas.1019645108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1019645108 PNAS Early Edition | 1of6 Downloaded by guest on October 2, 2021 and duplications that either encompassed an entire coding region or which produced an internal exon duplication. These genic CNVs were annotated for novelty and excluded from further consideration if they were found to have 5% or more overlap with any CNVs, either in the Database of Genomic Variants (http:// projects.tcag.ca/variation/) (27) or in a set of 3,000 control subjects not known to have CHD. We assessed specificity by attempting to confirm the 17 smallest novel CNVs from this set (CNVs encompassing 19 or fewer SNPs) by quantitative PCR; this set included seven deletions and 10 duplications. All but a single 10-SNP duplication were confirmed (Methods). fi Fig. 1. Anatomy in human heterotaxy. Situs solitus (SS). The cardiac apex is We identi ed 45 previously unrecorded genic CNVs in 39 oriented leftward, the right lung is trilobed, the left bilobed, the liver is on different subjects (Fig. 2 and Table S2). These CNVs included the right, and the stomach and spleen are on the left. Right atrial isomerism 16 heterozygous genic deletions, 25 complete duplications of at (RAI). Both lungs are trilobed, the liver is midline, and there is asplenia. least one gene, and four internal genic duplications. These CNVs Orientation of the cardiac apex is random, and complex CHD is found were in two size distributions: 38 were relatively small events, in >90% of affected patients. Left atrial isomerism (LAI). Both lungs are affecting one to five genes (27–1,488 kb, mean 2.1 genes per bilobed, the liver is midline, and there are multiple spleens. Orientation of CNV), and seven were larger chromosomal abnormalities, each the cardiac apex is random, and complex CHD is found in 80 to 90% of af- affecting > 90 genes (6–25 Mb, mean >250 genes per CNV). fected patients. Situs inversus (SI). Exact mirror-image of SS: the cardiac apex is rightward, there is a bilobed right lung and a trilobed left lung, the liver is Many more Htx cases than controls had rare genic CNVs [38 of 262 Htx subjects (14.5%) versus 73 of 991 controls (7.4%), P = on the left, and the stomach and spleen are on the right. − 1.5 × 10 4, ratio 2.0:1], consistent with CNVs playing a significant role in Htx development in some patients. To identify CNVs from SNP genotype intensities, we used We focused further evaluation on the 38 CNVs of smaller size, a likelihood-ratio based algorithm (26). In brief, quantitative anticipating that these may identify single genes with large effect intensity values of SNPs previously known to be present at 0, 1, 2, on Htx risk. None of the 61 genes altered by these CNVs had or 3 copies were used to determine the mean and SD of SNP been previously implicated in human Htx or any model of LR intensities for each class. We then determined the likelihood patterning. However, we did find genes in pathways previously ratio that a SNP with a given intensity has a copy number other linked to LR development, including 14 genes in either the cil- than 2. The likelihood ratios for consecutive SNPs are strung iary proteome, zinc-finger transcription factor family, or TGF-β together to assess the likelihood that segments of a given length signaling pathway (28–30). Remarkably, despite the small cohort represent a copy number other than 2. Applying this algorithm to size, we found that TGF-β receptor 2 (TGFBR2) was affected independent test data sets showed that 0 copy variants (homo- twice by unique opposite-state CNVs in two unrelated patients zygous deletions) can be highly specifically called by character- (independent deletion and duplication of TGFBR2) (Fig. 2A). istic intensity values with two or more consecutive SNPs, one copy variant (heterozygous deletions) by eight or more consec- Htx CNV Gene Expression Points to a Role in LR Development. To utive SNPs, and three copy variants (duplications) by 10 or more further analyze the 61 candidates, we undertook a screen in consecutive SNPs. Supporting data from B-allele frequencies X. tropicalis, a robust model for studying LR patterning. We used were required to make CNV calls (i.e., loss of heterozygosity for expression analysis to prioritize genes for investigation, focusing heterozygous deletions or shift of the nonhomozygous SNP B- on expression in the ciliated LR organizer [posterior notochordal allele fractions from 0.5 to 0.33 and 0.66 for duplications) (Fig. 2 plate or “node” in mouse, gastrocoel roof plate (GRP) in Xen- and Fig. S1). opus], the kidney (which has prominent cilia), and the cardiovas- To enrich for CNVs likely to have functional effect, we focused cular system. Of the 61 genes altered by novel CNVs, 22 had on genic CNVs, comprising deletions of at least one coding exon, orthologs in X. tropicalis, with high sequence similarity and iden-

Fig. 2. Rare genic deletions and duplications in Htx patients. Results of Illumina genotyping and qPCR are shown for four CNVs at three loci that contain genes implicated in Htx. In all panels, genes in the indicated chromosome segment are shown and genes implicated in Htx are denoted by arrows. Data from subjects with deletion or duplications are shown in red or blue, respectively, and the remaining Htx subjects are depicted in gray. (Upper) Probe intensities in consecutive 10-SNP windows normalized to a mean of 0 and SD of 1 from values in the remaining Htx subjects. P values supporting CNVs are shown in Table S1. Ratios of results of qPCR in index cases compared with controls are shown as red diamonds. (Lower) B-allele fraction (BAF) of SNPs across the interval. Arrows indicate locations of implicated LR genes. (A) Independent deletion and duplication affecting TGFBR2 in subjects 28 and 139, respectively. (B) Deletion of ﬁrst three exons of GALNT11 in subject 257. (C) Duplication of ROCK2 in subject 152.

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1019645108 Fakhro et al. Downloaded by guest on October 2, 2021 tical neighboring genes. We examined their developmental ex- xenopus.nibb.ac.jp) shows 40 with GRP expression; GRP-expressed pression patterns by whole-mount in situ hybridization (WMISH) genes are thus more than sevenfold enriched among rare Htx CNVs − at three developmental stages: neurula (stages 15–19, during (P = < 10 6). In addition, we identified 15 genes without Xenopus which symmetry at the GRP is first broken) (Fig. 3V), tailbud orthologs that had previous developmental expression studies in (stages 26–29, during which cardiac tube fusion occurs) (Fig. 3W), one or more vertebrate species, and none had noteworthy expres- and stages 33 to 36 (during which cardiac looping occurs) Fig. 3X). sion patterns. Inclusion of these 15 genes in the analysis of GRP/ fi WMISH was successful for 20 of these genes. Seven genes node enrichment in rare Htx CNVs continues to show signi cant × −4 exhibited nearly ubiquitous expression above that observed in enrichment (4.3-fold enrichment, P =2 10 ). A sense controls (Fig. S2 ), and the remaining 13 had more local- Knockdown in X. tropicalis Validates Htx Genes. Based on the sug- ized expression patterns (Fig. 3 and Fig. S2). Of these, seven genes gestive expression data of these seven genes, we examined the (tgfbr2, rock2, galnt11, lrrc8a, nek2, nup188, and greb1) had ex- effect of MO knockdown of these candidates on LR patterning. pression patterns consistent with a potential role in LR and car- This strategy was motivated by previous studies demonstrating diovascular development, with expression in at least two relevant dosage sensitivity for many chromosomal regions containing CHD domains: five of these localized to the GRP (Fig. 3 A, D, G, J, and genes, including: 22q11 (conotruncal defects) (31), 8p23 [AV M), two were prominently expressed in the early heart (Fig. 3 B, C, septal defects, Tetralogy of Fallot TOF)] (32), 7q11 (supravalvar E, and F), and five were expressed in the developing kidney (Fig. 3 aortic stenosis-del and PDA-dup) (33), 16p13 encompassing the H, I, L, N, O, Q, R, and T). GRP expression was highly enriched in CRBBP gene (Rubinstein-Taybi syndrome-del and septal defects- the Htx gene set. In comparison, WMISH of 845 genes (http:// dup) (34), and 1q21.1 (TOF) (35). Similarly, we identify independent patients with deletion and duplication affecting the region on chromosome 3 encompassing TGFBR2. Taken together, these data suggest that for many critical genes, either too much or too little gene activity results in CHD (36, 37). As positive controls for the MO experiments, we tested dnah9 and ift88, both previously reported to give robust LR phenotypes in Xenopus (38). As negative controls, we injected the standard negative control MO along with evaluation of dye-injected and uninjected controls. To address specificity of MO effects on LR phenotypes, we also examined 13 additional control genes: five MEDICAL SCIENCES genes from rare CNVs in Htx patients who did not have sug- gestive expression patterns (igfbp5, laptm5, runx2, smarcal1, tpk1, expression control) (Fig. S3), and eight genes from rare CNVs in controls (aldh1a1, cryzl1, gdf7, grik1, ccbl1, myh6, pdgfc, trat1, non-Htx control) (Fig. S3). All MOs were injected at the one-cell stage, and embryos were raised to stage 45/46, at which point heart- and gut-looping morphology were assessed. Because complete loss of function of some of these genes causes early embryonic lethality, we titrated doses of each MO from 0.5 to 20 ng per embryo and selected doses that minimally affected antero-posterior and dorso-ventral development for each. Final experimental doses for these stages ranged from 0.5 ng (60 fmol) to 4 ng (480 fmol) per embryo; standard control MO was injected at 8 ng per embryo. Heart looping was scored by established methods (21) as D-looped (normal) (Fig. 4A), L-looped (reversed) (Fig. 4C), or A-looped (midline/anterior-loop) (Fig. 4B). Gut looping was scored as either normal or abnormal (Fig. 4 D and E). At least 54 (mean n = 78) embryos were scored for each MO. Morphant embryos were scored by two independent readers blinded to group status, with 95% concordance of scoring for heart and gut phenotypes. Five of the seven Htx candidate MOs had striking effects on LR patterning, with abnormally looped hearts in 24 to 36% of embryos. The values in these five test morphants were equal to or greater than the effects of positive control MOs, and significantly greater than the standard negative control (all P ≤ 0.0001, for difference in heart looping) (Fig. 4G). Similarly, abnormal gut looping was seen with these same five MOs, with abnormal looping in 39 to 91% of embryos, again similar to positive controls and significantly greater than standard negative control (all P ≤ 0.0001) (Fig. 4H). Morphants of the two remaining test genes, greb1 and lrrc8a, gave no significant LR looping phenotype Fig. 3. WMISH analysis. (A–U) Results of in situ hybridization at three stages (Figs. S3 and S4). Notably, both greb1 and lrrc8a were in CNVs are shown for seven genes showing patterns of interest. These genes show that include another gene that gave a strong LR phenotype by expression in one or more of the following: GRP (blue arrows); heart or nup188 rock2, branchial arches (red arrows); kidney (green arrows). Stage 15 to 19 embryos MO knockdown ( and respectively) (Fig. S4, red are viewed dorso-posteriorly with anterior to the top to examine GRP ex- arrows). The absence of a strong LR phenotype in the greb1 and pression (shown schematically in V). Stage 26 to 29 and stage 33 to 36 em- lrrc8a morphants is consistent with a model in which a single bryos are viewed laterally with anterior to the left (shown schematically in W gene in these multiple gene CNVs contributes to Htx. These and X, respectively). striking effects on LR morphology showed specificity, because

Fakhro et al. PNAS Early Edition | 3of6 Downloaded by guest on October 2, 2021 none of the 13 additional control MOs had a significant effect on scored as showing left-sided, right-sided, bilateral, or absent pitx2 LR patterning (all P > 0.4 vs. standard negative control) (Fig. expression (Fig. 5). At least 45 (mean n = 74) embryos were S3). The difference in the frequency of LR defects between the scored in each group. Negative control morphants showed 88% test and control groups was highly significant (5/7 vs. 0/13, P = strong left-sided pitx2 expression, and 12% abnormal patterns. In 0.001 by Fisher’s exact test). contrast, consistent with prior reports (40), the positive control Htx can result from either a global effect on LR patterning or dnah9 morphants showed 35% abnormal pitx2 expression (P < − a heart-field–specific effect on cardiac looping. Because the five 2 × 10 4). All five test morphants that produced abnormal heart genes that showed abnormal LR patterning with MO knockdown and gut patterning also produced abnormalities in pitx2 expres- affected both heart and gut patterning, we expected that they sion (30–52% abnormality; all P < 0.002) (Fig. 5). should act upstream in the pathway. We tested this by examining Discussion expression of an early marker of global LR patterning, the transcription factor pitx2, which is normally induced by nodal We have identified rare genic CNVs in Htx patients and dem- signaling on the left side of the embryo, but inhibited on the right onstrated the effect of specific genes in these CNVs on LR de- side (39). We assayed pitx2 expression in morphants by WMISH velopment using MO knockdown in X. tropicalis, implicating at stages 28 to 30 (before heart-tube looping) using MO doses mutations in five genes. The evidence supporting phenotypic that minimally disrupted antero-posterior and dorso-ventral de- effect of these mutations includes highly significant enrichment velopment at this stage (0.5–8 ng per embryo). Embryos were of rare genic CNVs in Htx patients; highly significant enrichment in these CNVs of genes with expression in the GRP; and specific production of morphologic and molecular LR abnormalities by MO knockdown of these genes. Although the latter evidence indicates that modulation of these single genes is sufficient to produce LR phenotypes in Xenopus, we presume but cannot be certain that the same is true in humans. In addition, we found seven relatively large chromosomal abnormalities in these Htx patients, suggesting one or more genes in these CNVs also contribute to Htx. Included in this group was one 8p deletion

Fig. 4. LR abnormalities from MO knockdown in X. tropicalis. MOs were injected at the one-cell stage and heart and gut looping were assayed in tadpoles at stage 45/46. Views are from the ventral aspect, shown in schematic form in F.(A) Heart (area outlined in red box as in schematic in F) showing normal D-looping. The inflow (red arrow) is on the tadpole’s left, the outflow tract (yellow arrow) is on the tadpole’s right. (B) Heart showing abnormal, anterior, A-looping. The inflow (red) and outflow (yellow) are both at the midline, with no discernible LR orientation. (C) Heart showing abnormal, reversed, L-looping. The inflow (red) is on the tadpole’s right, the Fig. 5. Analysis of pitx2 expression in stage 28 to 30 Xenopus embryos. outflow (yellow) is on the tadpole’s left. (D) Normal clockwise rotation of the Embryos are viewed laterally from the left (first column) and the right gut. (E) Abnormal gut rotation. (F) Schematic of Xenopus tadpole at stage (second column). Note normal, bilateral pitx2 expression in the head region 45/46; ventral view with anterior to the top; arrows indicate heart and gut. in all embryos. (A) Expression of pitx2 is normally in the left lateral plate (G) Heart looping in MO knockdown tadpoles. Both dnah9 and ift88 are mesoderm (LPM, arrow). (B) Same normal embryo showing absent pitx2 positive controls; standard control MO (StdCtrl), uninjected control (UiC), expression in the right LPM. (C and D) Absent pitx2 expression. No pitx2 and dye-injected (Dye) are negative controls. Bars show the total percentage mRNA is found in the left or right LPM. (E and F) Bilateral pitx2 expression. of abnormally looped hearts: divided into A-loop (blue) and L-loop (red). (H) The pitx2 mRNA is found in both left and right LPM (arrows). (G and H) Right Gut looping in MO knockdown tadpoles. Both dnah9 and ift88 MOs are used pitx2 expression: pitx2 mRNA is absent from the left LPM, present in the as positive controls; StdCtrl, UiC, and Dye are used as negative controls. Red right LPM. (Graph) Summary of pitx2 mRNA expression in MO knockdown bars show the the percent of abnormal gut loops. Heart and gut looping embryos: dnah9 is a positive control; StdCtrl, and UiC are negative controls. were analyzed by two independent readers blinded to group status with Bars show the percent of abnormal pitx2 expression and are divided into − 95% concordance. *P < 10 4 vs. StdCtrl. blue (bilateral), red (right), and green (absent) LPM pitx2 expression.

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1019645108 Fakhro et al. Downloaded by guest on October 2, 2021 that included GATA4, which is known to result in diverse CHD GALNT11 has highly conserved domains similar to another family phenotypes (41). These findings collectively demonstrate that member, GALNT-like1, which has recently been shown to inhibit independent rare CNVs are likely related to pathogenesis in at nodal signaling by glycosylating activin receptor 1B, preventing least 3.8% of Htx patients; because there is no expression data for it from associating with activin receptor (55). This observation 26 of the genes in rare genic Htx CNVs, this number is likely suggests GALNT11 may play a related role in modulating TGF- underestimated. Although parental samples were not available β signaling. in this study, the demonstration that a high proportion of the The final gene implicated in our screen is NUP188, a compo- CNVs containing implicated genes were de novo events would nent of the nuclear pore complex. Nucleoporins function in strengthen the evidence for genetic causation and should be an transport of macromolecules between the nucleus and cytoplasm element of future studies. Rare CNVs have also been suggested to and in transcriptional regulation (56, 57). NUP188 is thought to play a role in diverse outflow tract lesions, including TOF (35). play a role in preventing the passage of integral membrane Because CNVs comprise a small fraction of mutation burden in proteins into the nucleus (58). How NUP188 functions in LR most genes, additional mutations with large effect will likely development remains to be elucidated. contribute to Htx, motivating efforts to identify them with tech- In summary, these findings support the notion that many ap- nologies such as whole-exome sequencing (26). parently sporadic cases of Htx have substantial contribution from It is noteworthy that all five genes identified in this study are rare genetic variation. They motivate further efforts to identify previously unrelated to LR patterning. This finding suggests that rare mutations in such patients. We believe the approach we prior studies of LR development have only uncovered a small have taken can be extended to other human developmental fraction of the genes required for LR patterning and may explain disorders, especially those for which morphology is conserved in why candidate gene resequencing has identified mutations in only a high-throughput system. We anticipate such studies will ad- 8 to 10% of Htx patients. In addition, diverse clinical phenotypes vance the understanding and treatment of Htx and other human within the Htx spectrum are seen among patients with these CNVs birth defects in the coming years. (Table S1). These phenotypes include transposition of the great arteries, abdominal situs inversus, asplenia/polysplenia, partial Methods anomalous pulmonary venous return, and aortic coarctation. Study Populations. We analyzed 262 heterotaxy patients (by genotype, 120 Among the five genes, NEK2 (never in mitosis kinase 2) (42) cluster with subjects of European ancestry, 104 Hispanic, 19 African Ameri- and ROCK2 (Rho-associated kinase 2) (43) are found in the cans, 19 all others). Patients with isolated D-andL-transposition of the great ciliary proteome. Cilia play a pivotal role in the earliest events in arteries were included. Coded DNA samples were obtained from two par- LR development, and ciliary defects underlie a disproportionate ticipating centers, Baylor College of Medicine (Houston, TX) and Yale Uni- MEDICAL SCIENCES number of laterality defects in mice, frogs, zebrafish, and humans versity School of Medicine (New Haven CT). The study was approved by the Yale and Baylor Institutional Review Boards. Patients were previously eval- (44). Mutations producing ciliary immotility cause primary ciliary uated for mutation in NODAL, ZIC3, CFC1, LEFTYA, LEFTYB, and ACVR2B. dyskinesia, often featuring Htx (12). Mutations expected to dis- Patients identified to have novel genic CNVs in our study did not have rupt centrosome function, ciliary biogenesis, or ciliary signaling, mutations in these genes. cause syndromes such as Bardet-Biedl syndrome (45) and Meckel- We also analyzed a control cohort comprising 999 subjects of European Gruber syndrome (46), featuring developmental defects including ancestry. We genotyped the controls on the same platform and subjected CHD. Nek2 is a member of a family of NIMA-related kinases them to the same QC steps and CNV analysis algorithms and parameters used implicated in cell cycle control and are defective in mouse cystic for CNV discovery in the patient cohort. None of the controls had congenital kidney disease, a known ciliary disorder (47). Knockdown of heart disease. Nek2 in mice results in failure to develop beyond the eight-cell stage (48). CNV Discovery and Novelty Assessment. Two-hundred eighty-eight Htx sam- fi β ples were genotyped on the Illumina 610Quad Beadchip platform (∼620,000 We also identi ed mutations in genes of the TGF- signaling ∼ pathway. Most notably, we found two subjects with mutations in SNP markers plus 60,000 CNP markers; average call rate: 99.77%, SD: 0.1%). β Of 288 Htx samples, 262 passed initial QC and were submitted for CNV dis- TGFBR2: one duplication and one deletion. The TGF- ligand covery. In the control cohort, 991 of 999 passed initial QC. CNVs were dis- nodal is critical in LR patterning (5); studies have largely focused covered using a likelihood ratio-based algorithm, as previously described on the nodal/activin branch of TGF-β signaling. Our findings (26), using thresholds of 8 and 10 consecutive probes for heterozygous implicate the TGF-β/TGFBR branch of this signaling network in deletions and duplications, respectively. In these high-confidence CNVs, we LR development, consistent with prior evidence that TGF-β1 assessed novelty by comparing their coordinates to known CNVs in the Da- overexpression in X. laevis resulted in LR defects (49). The finding tabase of Genomic Variants and CNVs called in 3,000 internal controls. CNVs of both a rare duplication and deletion of TGFBR2 suggests that were discarded if at least 5% of their length overlapped any CNV in these either increased or decreased signaling results in Htx, similar to two databases. Ninety-six CNVs passed this test and were considered high- fi other CHD genes (50). Point mutations in TGFBR2 cause Loeys- con dence unique variants. Dietz syndrome, which leads to aortopathy and a spectrum of fi Quantitative PCR Validation. For each tested CNV, at least two primer pairs CHD (51). Additionally, cardiac-speci c deletion of TGFBR2 in were designed within the boundaries of the CNV region. Additionally, a primer mice results in abnormal heart looping (52). Together with our set was designed to amplify a known copy-neutral segment of the ZNF423 data showing abnormal pitx2 expression in tgfbr2 morphants, this gene. The ratio of amplification of the test locus to amplification of the finding suggests that TGF-β signaling functions both in global ZNF423 diploid locus in patients and two sets of pooled controls were com- LR axis formation and at later stages in looping morphogenesis. pared in triplicate amplifications performed in parallel. In the absence of It is noteworthy that ROCK2 may provide a link between cilia deletion/duplication, the ratio of test locus:control locus in cases are expected and TGF-β signaling: in addition to being part of the ciliary pro- to be no different from the ratio in pooled controls (mean value 1.0). In teome, ROCK2 inhibits mesoderm induction in zebrafish em- contrast, these ratios should approximate 0.5 for heterozygous deletions and bryos by binding to and accelerating the lysosomal degradation of 1.5 for heterozygous duplications. Consistent results in triplicate samples with TGFBR1 (43). A general inhibitor of Rho kinases was shown to at least two primer pairs were required to declare a conclusive result. Quan- titative PCR was performed on 23 loci (14 duplications, 9 deletions), 17 of affect LR development in chick (53), and after submission of this < fi which were the smallest ( 19 probes) rare genic CNVs. All but one 10-probe article, rock2b was implicated in LR patterning in zebra sh by duplication produced qPCR results concordant with the copy-number state MO knockdown, supporting the evolutionary conservation of this predicted by our CNV detection algorithm (>95% specificity). gene’s role in LR patterning (54). GALNT11, which encodes a glycosyl transferase, was also Xenopus Analysis. X. tropicalis orthologs of human genes were identified identified in this screen. This finding is noteworthy in that using Metazome (www.metazome.com) or Xenbase (www.xenbase.org)

Fakhro et al. PNAS Early Edition | 5of6 Downloaded by guest on October 2, 2021 Web sites. In each case, evidence of synteny to the human genome was Statistical Analysis. Statistical comparisons between groups were by χ2 sta- required to conﬁrm orthology. One gene, CETN1, is a member of a highly tistics unless expected cell frequencies were less than 5, in which case Fisher’s homologous multigene family and was not investigated further. WMISH was exact test was used. performed as previously described (24). We identiﬁed GRP expression in control genes from a public database (http://xenopus.nibb.ac.jp/). MOs were ACKNOWLEDGMENTS. We thank the patients and families who participated injected at the one-cell stage and embryos scored at stage 45/46 for LR in this study and the staff of the Yale West Campus Center for Genomic phenotypes and stages 26 to 29 for pitx2. Clones used for generating anti- Analysis for their contributions to this project, Carol Nelson-Williams for sense probes are available upon request. Xenopus procedures were reviewed helpful discussions and advice, and Michael Slocum and Sarah Kirschner for and approved by Yale’s Institutional Animal Care and Use Committee, which animal husbandry. This study was supported by National Institutes of Health is Association for Assessment and Accreditation of Laboratory Animal Care- Grants R01HD045789 and R01HL093280 (to M.B.) and R01DE018824 and accredited. Additional detailed methods and a summary of MO sequences R01DE018825 (to M.K.K.). R.P.L. is an Investigator of The Howard Hughes and doses are available in SI Methods and Table S3. Medical Institute.

1. Pierpont ME, et al.; American Heart Association Congenital Cardiac Defects 29. Ostrowski LE, et al. (2002) A proteomic analysis of human cilia: Identification of novel Committee, Council on Cardiovascular Disease in the Young (2007) Genetic basis for components. Mol Cell Proteomics 1:451–465. congenital heart defects: Current knowledge: A scientific statement from the 30. Pazour GJ, Agrin N, Leszyk J, Witman GB (2005) Proteomic analysis of a eukaryotic American Heart Association Congenital Cardiac Defects Committee, Council on cilium. J Cell Biol 170:103–113. Cardiovascular Disease in the Young: Endorsed by the American Academy of 31. Portnoï MF (2009) Microduplication 22q11.2: A new chromosomal syndrome. Eur J Pediatrics. Circulation 115:3015–3038. Med Genet 52:88–93. 2. Cohen MS, et al. (2007) Controversies, genetics, diagnostic assessment, and outcomes 32. Barber JC, et al. (2008) 8p23.1 duplication syndrome; a novel genomic condition with relating to the heterotaxy syndrome. Cardiol Young 17(Suppl 2):29–43. unexpected complexity revealed by array CGH. Eur J Hum Genet 16:18–27. 3. Taketazu M, Lougheed J, Yoo SJ, Lim JS, Hornberger LK (2006) Spectrum of 33. Van der Aa N, et al. (2009) Fourteen new cases contribute to the characterization of cardiovascular disease, accuracy of diagnosis, and outcome in fetal heterotaxy the 7q11.23 microduplication syndrome. Eur J Med Genet 52:94–100. syndrome. Am J Cardiol 97:720–724. 34. Thienpont B, et al. (2010) Duplications of the critical Rubinstein-Taybi deletion region 4. Levin M, Johnson RL, Stern CD, Kuehn M, Tabin C (1995) A molecular pathway on chromosome 16p13.3 cause a novel recognisable syndrome. J Med Genet 47: determining left-right asymmetry in chick embryogenesis. Cell 82:803–814. 155–161. 5. Shiratori H, Hamada H (2006) The left-right axis in the mouse: From origin to 35. Greenway SC, et al. (2009) De novo copy number variants identify new genes and loci morphology. Development 133:2095–2104. in isolated sporadic tetralogy of Fallot. Nat Genet 41:931–935. 6. Bisgrove BW, Morelli SH, Yost HJ (2003) Genetics of human laterality disorders: 36. Mefford HC, et al. (2008) Recurrent rearrangements of chromosome 1q21.1 and – Insights from vertebrate model systems. Annu Rev Genomics Hum Genet 4:1 32. variable pediatric phenotypes. N Engl J Med 359:1685–1699. 7. Nonaka S, et al. (1998) Randomization of left-right asymmetry due to loss of nodal 37. Kontaridis MI, Swanson KD, David FS, Barford D, Neel BG (2006) PTPN11 (Shp2) fl fl cilia generating leftward ow of extraembryonic uid in mice lacking KIF3B motor mutations in LEOPARD syndrome have dominant negative, not activating, effects. – protein. Cell 95:829 837. J Biol Chem 281:6785–6792. 8. McGrath J, Somlo S, Makova S, Tian X, Brueckner M (2003) Two populations of node 38. Schweickert A, et al. (2010) The nodal inhibitor coco is a critical target of leftward – monocilia initiate left-right asymmetry in the mouse. Cell 114:61 73. flow in Xenopus. Curr Biol 20:738–743. 9. Bartoloni L, et al. (2002) Mutations in the DNAH11 (axonemal heavy chain dynein 39. Piedra ME, Icardo JM, Albajar M, Rodriguez-Rey JC, Ros MA (1998) Pitx2 participates type 11) gene cause one form of situs inversus totalis and most likely primary ciliary in the late phase of the pathway controlling left-right asymmetry. Cell 94:319–324. – dyskinesia. Proc Natl Acad Sci USA 99:10282 10286. 40. Vick P, et al. (2009) Flow on the right side of the gastrocoel roof plate is dispensable 10. Gebbia M, et al. (1997) X-linked situs abnormalities result from mutations in ZIC3. Nat for symmetry breakage in the frog Xenopus laevis. Dev Biol 331:281–291. – Genet 17:305 308. 41. Tomita-Mitchell A, Maslen CL, Morris CD, Garg V, Goldmuntz E (2007) GATA4 11. Karkera JD, et al. (2007) Loss-of-function mutations in growth differentiation factor-1 sequence variants in patients with congenital heart disease. J Med Genet 44:779–783. (GDF1) are associated with congenital heart defects in humans. Am J Hum Genet 81: 42. Fry AM (2002) The Nek2 protein kinase: A novel regulator of centrosome structure. 987–994. Oncogene 21:6184–6194. 12. Kennedy MP, et al. (2007) Congenital heart disease and other heterotaxic defects in 43. Zhang Y, et al. (2009) Rock2 controls TGFbeta signaling and inhibits mesoderm a large cohort of patients with primary ciliary dyskinesia. Circulation 115:2814–2821. induction in zebrafish embryos. J Cell Sci 122:2197–2207. 13. Kosaki K, et al. (1999) Characterization and mutation analysis of human LEFTY A and 44. Eggenschwiler JT, Anderson KV (2007) Cilia and developmental signaling. Annu Rev LEFTY B, homologues of murine genes implicated in left-right axis development. Am J Cell Dev Biol 23:345–373. Hum Genet 64:712–721. 45. Nachury MV, et al. (2007) A core complex of BBS proteins cooperates with the GTPase 14. Kosaki R, et al. (1999) Left-right axis malformations associated with mutations in Rab8 to promote ciliary membrane biogenesis. Cell 129:1201–1213. ACVR2B, the gene for human activin receptor type IIB. Am J Med Genet 82:70–76. 46. Dawe HR, et al. (2007) The Meckel-Gruber Syndrome proteins MKS1 and meckelin 15. Roessler E, et al. (2009) Cumulative ligand activity of NODAL mutations and modifiers interact and are required for primary cilium formation. Hum Mol Genet 16:173–186. are linked to human heart defects and holoprosencephaly. Mol Genet Metab 98: 47. Mahjoub MR, Trapp ML, Quarmby LM (2005) NIMA-related kinases defective in 225–234. murine models of polycystic kidney diseases localize to primary cilia and centrosomes. 16. Schwabe GC, et al. (2008) Primary ciliary dyskinesia associated with normal axoneme J Am Soc Nephrol 16:3485–3489. ultrastructure is caused by DNAH11 mutations. Hum Mutat 29:289–298. 48. Sonn S, Khang I, Kim K, Rhee K (2004) Suppression of Nek2A in mouse early embryos 17. van Bon BW, et al. (2008) Transposition of the great vessels in a patient with a 2.9 Mb confirms its requirement for chromosome segregation. J Cell Sci 117:5557–5566. interstitial deletion of 9q31.1 encompassing the inversin gene: clinical report and 49. Mogi K, et al. (2003) Xenopus neurula left-right asymmetry is respeficied by review. Am J Med Genet A 146A:1225–1229. – 18. Sutherland MJ, Ware SM (2009) Disorders of left-right asymmetry: Heterotaxy and microinjecting TGF-beta5 protein. Int J Dev Biol 47:15 29. situs inversus. Am J Med Genet C Semin Med Genet 151C:307–317. 50. Laitenberger G, Donner B, Gebauer J, Hoehn T (2008) D-transposition of the great – 19. Conrad DF, et al.; Wellcome Trust Case Control Consortium (2010) Origins and arteries in a case of microduplication 22q11.2. Pediatr Cardiol 29:1104 1106. functional impact of copy number variation in the human genome. Nature 464: 51. Loeys BL, et al. (2005) A syndrome of altered cardiovascular, craniofacial, neuro- 704–712. cognitive and skeletal development caused by mutations in TGFBR1 or TGFBR2. Nat – 20. Craddock N, et al.; Wellcome Trust Case Control Consortium (2010) Genome-wide Genet 37:275 281. association study of CNVs in 16,000 cases of eight common diseases and 3,000 shared 52. Jiao K, et al. (2006) Tgfbeta signaling is required for atrioventricular cushion controls. Nature 464:713–720. mesenchyme remodeling during in vivo cardiac development. Development 133: – 21. Blum M, et al. (2009) Xenopus, an ideal model system to study vertebrate left-right 4585 4593. asymmetry. Dev Dyn 238:1215–1225. 53. Wei L, et al. (2001) Rho kinases play an obligatory role in vertebrate embryonic – 22. Schweickert A, et al. (2007) Cilia-driven leftward flow determines laterality in organogenesis. Development 128:2953 2962. Xenopus. Curr Biol 17:60–66. 54. Wang G, et al. (2011) The Rho kinase Rock2b establishes anteroposterior asymmetry 23. Hellsten U, et al. (2010) The genome of the Western clawed frog Xenopus tropicalis. of the ciliated Kupffer’s vesicle in zebrafish. Development 138:45–54. Science 328:633–636. 55. Herr P, Korniychuk G, Yamamoto Y, Grubisic K, Oelgeschläger M (2008) Regulation of 24. Khokha MK, et al. (2002) Techniques and probes for the study of Xenopus tropicalis TGF-(beta) signalling by N-acetylgalactosaminyltransferase-like 1. Development 135: development. Dev Dyn 225:499–510. 1813–1822. 25. De Luca A, et al. (2010) Familial transposition of the great arteries caused by multiple 56. Kalverda B, Pickersgill H, Shloma VV, Fornerod M (2010) Nucleoporins directly mutations in laterality genes. Heart 96:673–677. stimulate expression of developmental and cell-cycle genes inside the nucleoplasm. 26. Choi M, et al. (2009) Genetic diagnosis by whole exome capture and massively parallel Cell 140:360–371. DNA sequencing. Proc Natl Acad Sci USA 106:19096–19101. 57. Capelson M, et al. (2010) Chromatin-bound nuclear pore components regulate gene 27. Iafrate AJ, et al. (2004) Detection of large-scale variation in the human genome. Nat expression in higher eukaryotes. Cell 140:372–383. Genet 36:949–951. 58. Theerthagiri G, Eisenhardt N, Schwarz H, Antonin W (2010) The nucleoporin Nup188 28. Avidor-Reiss T, et al. (2004) Decoding cilia function: Defining specialized genes controls passage of membrane proteins across the nuclear pore complex. J Cell Biol required for compartmentalized cilia biogenesis. Cell 117:527–539. 189:1129–1142.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1019645108 Fakhro et al. Downloaded by guest on October 2, 2021